No abstract

Carcinosarcoma is an aggressive neoplasm of the female genital tract, which comprises 1-2% of malignancies of the uterine corpus. Because of the broad range of differentiation exhibited by these tumors, the precise nature of the relationship between epithelial and stromal components in this unique tumor remain unclear. Previous studies have demonstrated that mutation and consequent overexpression of the tumor suppressor gene p53 occurs frequently in carcinosarcoma and is conserved from primary to metastastic sites. We examined p53 accumulation in formalin-fixed, paraffin-embedded archival sections in 19 cases previously shown to have mutations in the p53 gene and performed semi-quantitative analysis of the intensity of staining and relative density of positive cells and stromal and glandular elements. There was a high level of concordance of immunohistochemical staining for the p53 oncoprotein between glandular and stromal elements. These results further suggest a clonal origin for the diverse elements of carcinosarcoma.

Electrical brain stimulation is a proven therapy for epilepsy, but long-term seizure free outcomes are rare. Early implantable devices were developed for open-loop stimulation without sensing, embedded computing or adaptive therapy. Recent device advances include sensing and closed-loop responsive stimulation, but these clinically available devices lack adequate computing, data storage and patient interface to concisely catalog behavior, seizures, and brain electrophysiology, despite the critical importance of these details for epilepsy management. Here we describe the first application of a distributed brain co-processor providing an intuitive, bi-directional interface between device implant, patient & physician, and implement it in human and canine patients with epilepsy living in their natural environments. Automated behavioral state tracking (awake and sleep) and electrophysiologic classifiers for interictal epileptiform discharges and electrographic seizures are run on local hand-held and distributed cloud computing resources to guide adaptive electrical stimulation. These algorithms were first developed and parameterized using long-term retrospective data from 10 humans and 11 canines with epilepsy and then implemented prospectively in two pet canines and one human with drug resistant epilepsy as they naturally navigate their lives in society.